All I want is an ordinary RDBMS with SPAAS (Stored Procedures as a Service) that return a binary blob of tables, nothing else. I don't want noSQL, I don't care about cloud objects, I saw Firebase and didn't like it, I don't want JSON as a result, I don't care if noobs don't know who CODD is, they will learn it the hard way sooner or later.

Oracle is the living god of RDBMS and they should take advantage of that yet they don't. It has nothing to do with AI, ML, NN, or any other word du jour, just plain tables lightning fast.

I almost do this. My main APIs are POST/GET sqlite databases (ie: I build a sqlite db and return it like a file, then in clients and use it. Not using stupid JSON when I need long rows!).

--

I'm dreaming about building a relational language and by the way build a RDBMS on top of maybe sqlite. I think exist a lot of potential for a relational store with more flexibility

Think: with both in-memory and on-disk data storage, fixed schema (as normal) + dynamic (like JSON) + Flexible.

By flexible I mean to use Union Types and return mixed rows:

//A union of several schemas: Table Log = | Customer | Order | OrderLine

SELECT Log =

Customer(1,...). Customer(2,...) Order(1, ...) OrderLine(1, ...)

Also, maybe allow to directly acces indexes (so a index is just a KV store!), to allow logical log manipulation, etc.

In short, what about a relational-redis?

Transferring whole sqlite databases makes sense when one side (usually the client) really uses the file as semi-persistent database for some non-trivial timespan (we actually do that for most of our mobile applications, usually for long-lived read only reference data, but one of our applications actually POSTs modified sqlite files back to server as part of normal operation, which is in fact mainly a hackish way to overcome memory limitations of the hardware involved). Using it only as data format for the transfer is hugely inefficient in terms of both time and bandwidth.

Yes. That is why a better options is to have a kind of streamable relational format (a JSON for tables ?).

The problem is that using sqlite it also embed a functional query engine, when a simple data format requiere extra layers to make it work.

So even better is to have a sqlite-like engine that allow partial/streamable reading/writes.

You can pay for managed hosting for commercial databases. What further requirements are you missing?

>> All I want is an ordinary RDBMS

It doesn't matter what you want. Humans are no longer the customers; we are now the products and the slaves of machines.

At least that's what the article makes it sound like.

Think they are looking for the next thing with #18c cause the are so late to the game they cant just be a faster AWS Me2

that wont do it

Yaniv, they just leapfrogged everyone back into the game with an obvious large investment of R&D. Who invests in R&D these days anyway?

> Who invests in R&D these days anyway?

Google, Microsoft and Amazon invest billions _per quarter_ in R&D.

Totally agree! just make it work fast, I'll take care of the rest

> Oracle will have a steep climb to pull off their vision. They eventually will... someday. It's not with the fortune 100s that you surpass biological evolution, it's with the hackers, startups, developers, and red-eyed bandits that code at 3:14 AM to solve that one bug that kept them from sleeping for the past 3 days. The Revolution of Evolution in cloud ML and autonomous DB will come from a Panoply, the execution of large enterprise market adoption will come from an Oracle.

Lol sounds like a great place to work.

> red-eyed bandits that code at 3:14 AM to solve that one bug that kept them from sleeping for the past 3 days

I don't want to apply a fix written at 3:14AM by a developer with sleep deprivation. That's the same as being proud that drunk developers fix the most critical bugs.

It's amazing how freeing working only from 9 to 5 and not having access to emails away from the office is. I'm surprised anyone would willingly work that crazy

I don't even get why they would. It's not like the time I have to type code is my limiting factor on productivity.

Past a certain point, I'm not solving issues -- I'm creating them. On any horizon over a week, I'm better off sleeping, eating, exercising, etc than trying to smash every hour into working. Keeping my brain on a routine and properly nurtured is more productive than crazy hours, by a wide margin.

But I guess looking busy can be as important as results.

So let the one who thinks he is standing beware that he does not fall. This is a biblical reference, but in this case, it applies. He has built up this product to do a lot by attempting to step on the neck of AWS. He's barely reached an elbow. What he is attempting to start is already being done. http://blog.panoply.io/the-power-struggle-in-the-cloud-db-wa...

When is someone going to make this:

Example 1

---------

Me: "Create a new Customer Record"

DB: "Customer Entity doesn't exist, shall I create it?"

Me: "Yes"

DB: "Customer Entity created. What is the Name of your Record?"

Me: "Larry Ellison"

DB: "Customer, Larry Ellison created"

Example 2

---------

Me: "How many customers are there?"

DB: "There are 123 customers"

Example 3

---------

Me: "How many customers have accounts that are past due"

DB: "I did not understand past due"

Me: "How many customers have accounts where past due is true?"

DB: "There are 5 customers where account past due is true"

Me: "Send me an email of those customers"

DB: "Email has been sent"

It could be a conversational API layer for any underlying database. Intriguing idea, possible with existing technologies.. Tune in next week for ChatQL!

Check out http://kueri.me/ it wont do everything in your examples

FriendlyData already made this: https://friendlydata.io/

Some of this functionality (using NLP queries) is built into Microsoft PowerBI. It's not fully there yet but I think the idea is to make it easy to query databases.

Microsoft's PowerBI product is trying to push out a natural language query processor like in your examples. Have you had a chance to try it out yet?

That is called programming. It works better with a simple language instead of English.

The asking questions side is in Google sheets (and analytics recently).

Sounds interesting! Looks like Siri on a DB.

It's made. It's called a person with a 3 year college degree.

One day AI will be smarter than a college degree, but it's a while off.

If you want a simple subset there's heaps of CRMs.

Wouldn't a much bigger deal be distributed part of the Oracle database in the cloud, not ML? Because turning a feature rich RDBMS into a decent distributed database is not really possible. You either have to sacrifice consistency, which by the looks of it is what Oracle is thinking to do [1], or you have to sacrifice both performance and availability. And there are already cheap not CAP consistent RDBMSs as services in the cloud. It's just hard to imagine how Oracle can leverage 90s era technology in a distributed world.

[1] http://oracle-help.com/articles/oracle-18c-future-database/

I f the reason they want to distribute was performance then as a DB performance engineer, I can tell you first hand that tuning a database, indexing and caching can keep 99.9% of the internet companies very happy.

I know its hard to imagine, but it is true and most devops teams look to a shiny new tech they saw at a meet up than (like you said) fix 90s era technologies.

Note Google Spanner.

Yes, Spanner is the opposite of that strategy, designed from the ground up as a distributed database.

And CockroachDB, TiDB, etc...

I'm not quite sure I can be hyped. It's worth keeping an eye on. The core dbms is relatively easy to carry and with some careful considerations can be formed and firmed up quite nicely. The irony is the that thicker the oracle kit the more intricate, delicate and expensive it becomes. Storage, db engine, memory and.. cluster ware? Nope, that bird can-nah fly.

Nice database you got there, it'd be a shame if it couldn't grow. Eight figures later.

If ml can expect, select and plant my data closer to the end user then my core rdbms engine can grow, otherwise you're just moving pieces on the monopoly board. Fat pipes, willing victims and a little ingenuity can go a long way.

Moving infrastructure to the cloud can save money and labor, but the value of working with an experienced DBA to me personally has been getting expert feedback on schema design and stored procedure and query performance.

Debating Oracle and the cloud always seems a bit of to me. I believe that most of Oracles database customers specifically don't want a cloud solution.

Ellison has always been a visionary, his problem is getting his programmers to actually act and make that vision come true.

#18c sounds bold

ML DB lightyears ahead? More like what everyone is doing. THis author needs to catchup on whats going on at CMU with Andy Palvo. PG will run this way before ORacle

How can we even know #18c will do anything? it sounded like a bunch of buzzwords!

Oracle is so late to the game its gonna take a miracle for them to make a dent in AWS

Reading about 18c it sounds like the focus will be on applying ML to performance tuning the database, this is actually a pretty solid thing to pursue as performance tuning will usually have dramatic results and it's a knowledge area that is isolated and esoteric.

Larry Ellison never bullshits, the question is if its actually gonna as autonomous as he says. From my experience with oracle DBs its gonna do half of what he claims and cost twice what he claims

If it doesn't do what he says or cost what he says it costs then how is that not BS? Did I miss the joke?

Oracle has a lot of "legacy" customers who have a lot invested in their existing Oracle infrastructure. Oracle wants to migrate these customers to their cloud. For "Enterprise" I see Oracle's biggest competitor being Azure.

Yes...I agree. I don't think this is a fast death. But it is similar to the position all the RISC Unix vendors were in once X64 came out. Took a decade, but that was the death knell.

This one might be longer, say 15 or 20 years...but it's coming.

They might need a new database product to survive if they are going to have to compete fairly. Maybe acquire a bunch of distributed systems startups, like CockroachDB, and do something with them for the Oracle cloud.

I think you are reading this wrong. Imaging you are a technical co-founder of a new start up... Someone is now telling you: 'you will not need to hire as many devops or DBAs if you buy this technology and you are also future proof, because we tune things for you and scale up your hardware without any effort from you.'

If the price is good (and considering you would not have to pay for salaries as well) you would be a fool not to take it.

>If the price is good

That's not Oracle's historical strong suit.

Sure, but their non database features probably aren't in the same league as the top 3.

They actually have some features that aren't available in other clouds (well, AWS/GCP/AZ) -- if we weren't so locked into AWS, we might've tried them for that reason. We wanted to know if we could run our own hypervisor in their cloud and access VT-x instructions. (And there were other companies at their talk that were definitely hitting up sales for trials; the "baremetal" aspect intrigued a lot of people with compliance reqs or complex cloud appliances. Dedicated bandwidth was another.)

I think later clouds have advantages and not just missing/worse features -- they got to design their architecture after seeing others try, and definitely learned some lessons.